The crystal quality of 2201 and 2212 films grown on a SrTiO3 substrate by ALL-MBE is very high, as described above. The only defects which are commonly observed in these films are twin boundaries, one of which is shown in Figure 5. Part of the image has the 2212 a-axis oriented normal to the plane of the paper; this portion appears to have long, unbroken BiO planes. The other part of the image has the 2212 b-axis oriented normal. The b-axis-oriented portion of the image displays the incommensurate modulation which has previously been observed in 2212 films. It has been shown that the growth of untwinned films via ALL-MBE is possible by using specially miscut substrates. However, the special precautions necessary to prevent twinning are not deemed to be worthwhile for routine growths since the presence of twins is not believed to affect the performance of Josephson junctions.
Occasionally more troublesome defects may occur. These defects include stacking faults and antiphase boundaries, as Figure 6 illustrates. The stacking fault defects are most easily visualized by observing the presence of the wrong number of BiO planes, although presumably the SrO planes and CaCuO2 planes may also be affected. There are several places in the image of Figure 6 where apparently more than two BiO planes occur together. Nearby in the image, an arrow points to the spot where layer spacings indicate that two phases of different stoichiometry have grown together. Estimates of the layer thicknesses for these phases indicate that they may correspond to 2223, which has an extra CaCuO2 slab, and 3312, which has an extra BiO-SrO unit. Exactly why these off-stoichiometric phases and stacking fault defects occur has not yet been determined. For reasons which are not well-understood, the frequency of occurrence of these defects is greater above a 2278 or 1278 barrier than in a 2212 film grown on a 2201 buffer layer. It may be hypothesized that the unfavorable charge balance in the unstable barrrier layer may cause havoc in subsequent overgrowths.
RBS, EDAX and SIMS measurements indicate that the overall composition of the films is within a few percent of the nominal stoichiometry. However, defects which are visible on the surface of the superconducting films with optical microscopy are found by RBS, SIMS and EDAX to have a substantially different stoichiometry. These surface defects typically occur in the corners of the films, where the deviation from the nominal stoichiometry would be expected to be greatest because the thermal effusion sources are not exactly on-axis. A correlation between the occurrence of these defects and poor Josephson junction performance has been observed. Whether any of the defects which are visible in plan view with optical microscopy are identifiable with the defects seen in TEM cross-sections is not yet clear. The length scale of the the stacking faults and antiphase boundaries which are observed in cross-section is angstroms, while the size of optically observable surface defects is microns. Nonetheless, it may be that the nanoscale imperfections in film structure serve as nucleation points for the larger defects which are observed in plan view.
Despite the presence of planar defects and twins in the BiSrCaCuO films, the overall impression that TEM observations leave is one of a high degree of ordering. The low-magnification TEM image shown in Figure 7 illustrates this point. The bright layers in the image are three barrier layers, each designed to consist of a single molecular layer of 2278. Clearly these barriers are continuous over very large in-plane lengths, perhaps 1000 Å. The in-plane continuity of the barriers is clearly larger than that out-of-plane, once again illustrating the inherent anisotropy of BiSrCaCuO growth. An estimate of the average film non-stoichiometry can be made from an enlarged version of the image in Figure 7 by comparing the observed total thickness of the film to the thickness of a perfect film with the same layering sequence. The comparison shows that the Bi:Cu ratio of the film deviates from the target composition by only about 1%. The thickness is determined by the Bi:Cu ratio since it is this ratio that determines which of the BiSrCaCuO compounds is formed.